Unsupervised domain adaptation for video classification

1. A computer-implemented method for unsupervised domain adaptation for video classification, comprising:

learning, by a hardware processor, a respective transformation for each of a plurality of target video clips taken from a set of target videos, responsive to original features extracted from the plurality of target video clips, the respective transformation for correcting differences between a target domain corresponding to the plurality of target video clips and a source domain corresponding to a plurality of source video clips taken from a set of source videos;

adapting, by the hardware processor, the target domain to the source domain by applying the respective transformation to the original features extracted from the plurality of target video clips to obtain transformed features for the plurality of target video clips;

converting, by the hardware processor, the original features and the transformed features of same ones of the plurality of target video clips into a single classification feature for each of the target videos in the set; and

classifying, by the hardware processor, a new target video relative to the set of source videos using the single classification feature for each of the target videos in the set,

wherein said learning step learns the respective transformation that corrects for motion differences between the source domain and the target domain.

2. The computer-implemented method of claim 1 , further comprising extracting the original features using a convolutional neural network.

3. The computer-implemented method of claim 1 , wherein said learning step learns the respective transformation that corrects for viewpoint differences between the source domain and the target domain.

4. The computer-implemented method of claim 1 , further comprising performing source to target video alignment for said classifying step based on a domain adversarial loss which predicts which of the features are from the source domain and which of the features are from the target domain.

5. The computer-implemented method of claim 1 , wherein said converting step is performed on the original features and the transformed features using a dimension-wise maximum function.

6. The computer-implemented method of claim 1 , wherein said converting step is performed the original features and the transformed features using an averaging function.

7. The computer-implemented method of claim 1 , wherein said classifying step selectively uses a classification loss or a verification loss responsive to source and target labels being identical or different, respectively.

8. The computer-implemented method of claim 1 , further comprising using a classification loss in said classifying step responsive to source and target labels being identical, the source and target labels corresponding to a particular one of the plurality of source video clips and a particular one of the plurality of target video clips, respectively.

9. The computer-implemented method of claim 1 , further comprising using a verification loss in said classifying step responsive to source and target labels being different, the source and target labels corresponding to a particular one of the plurality of source video clips and a particular one of the plurality of target video clips, respectively.

10. The computer-implemented method of claim 1 , wherein said classifying step comprising using a feature embedding for the new target video relative to the target domain for a near-neighbor based classification.

11. The computer-implemented method of claim 1 , wherein the hardware processor uses a Support Vector Regressor to predict parameters of the respective transformation.

12. The computer-implemented method of claim 1 , further comprising capturing, by a drone, the target videos in the set.

13. A computer program product for unsupervised domain adaptation for video classification, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:

learning, by a hardware processor, a respective transformation for each of a plurality of target video clips taken from a set of target videos, responsive to original features extracted from the plurality of target video clips, the respective transformation for correcting differences between a target domain corresponding to the plurality of target video clips and a source domain corresponding to a plurality of source video clips taken from a set of source videos;

adapting, by the hardware processor, the target domain to the source domain by applying the respective transformation to the original features extracted from the plurality of target video clips to obtain transformed features for the plurality of target video clips;

converting, by the hardware processor, the original features and the transformed features of same ones of the plurality of target video clips into a single classification feature for each of the target videos in the set; and

classifying, by the hardware processor, a new target video relative to the set of source videos using the single classification feature for each of the target videos in the set,

wherein said learning step learns the respective transformation that corrects for motion differences between the source domain and the target domain.

14. The computer-implemented method of claim 13 , wherein said learning step learns the respective transformation that corrects for viewpoint differences between the source domain and the target domain.

15. The computer-implemented method of claim 13 , wherein the method further comprises performing source to target video alignment for said classifying step based on a domain adversarial loss which predicts which of the features are from the source domain and which of the features are from the target domain.

16. The computer-implemented method of claim 13 , wherein said converting step is performed on the original features and the transformed features using a dimension-wise maximum function.

17. The computer-implemented method of claim 13 , wherein said converting step is performed the original features and the transformed features using an averaging function.

18. A computer processing system for unsupervised domain adaptation for video classification, comprising:

a memory device including program code stored thereon; and

a processor, operatively coupled to the memory device, and configured to run the program code stored on the memory device to

learn a respective transformation for each of a plurality of target video clips taken from a set of target videos, responsive to original features extracted from the plurality of target video clips, the respective transformation for correcting differences between a target domain corresponding to the plurality of target video clips and a source domain corresponding to a plurality of source video clips taken from a set of source videos;

adapt the target domain to the source domain by applying the respective transformation to the original features extracted from the plurality of target video clips to obtain transformed features for the plurality of target video clips;

convert the original features and the transformed features of same ones of the plurality of target video clips into a single classification feature for each of the target videos in the set; and

classify a new target video relative to the set of source videos using the single classification feature for each of the target videos in the set,

wherein said learning step learns the respective transformation that corrects for motion differences between the source domain and the target domain.